CN-122016841-A - Contaminant detection method, device, detector and storage medium

CN122016841ACN 122016841 ACN122016841 ACN 122016841ACN-122016841-A

Abstract

The application discloses a pollutant detection method, a device, a detector and a storage medium, and relates to the technical field of device detection, wherein the method is applied to the detector provided with a multispectral illumination component and an imaging component; and identifying the current spectral vector through a preset pollution detection model to obtain a pollution detection result, wherein the preset pollution detection model is a deep learning model obtained based on sample spectral image training of various pollutants. Compared with the existing single-wavelength visual detection, the method can quickly identify the components of the pollutants by using the trained preset pollution detection model in the on-site adjustment link, and provides a basis for accurate cleaning operation.

Inventors

CHEN TAIXI
SU HANG
ZHANG XIN
LI XIN
ZHANG YINGHUAI
FENG CHANG

Assignees

季华实验室

Dates

Publication Date: 20260512
Application Date: 20251222

Claims (10)

1. A contaminant detection method, the contaminant detection method being applied to a detector provided with a multispectral illumination component and an imaging component, the method comprising: the imaging component is used for acquiring images of the illuminated optical element to be detected to obtain a multi-band spectrum image, and the multi-spectrum illumination component is used for illuminating the optical element to be detected placed in the detector according to multi-spectrum bands; extracting features of the spectrum image to obtain a current spectrum vector corresponding to the spectrum image; And identifying the current spectrum vector through a preset pollution detection model to obtain a pollution detection result, wherein the preset pollution detection model is a deep learning model based on sample spectrum image training of various pollutants.
2. The method of claim 1, wherein the step of extracting features from the spectral image to obtain a current spectral vector corresponding to the spectral image comprises: Performing flat field correction on the spectrum image according to a preset whiteboard image to obtain a corrected spectrum image, wherein the whiteboard image is a reference image obtained by photographing a diffuse reflection standard whiteboard; extracting the relative reflectivity value of each pixel point by taking each pixel point in the corrected spectrum image as a unit; And arranging the relative reflectivity values of each pixel point according to the wavelength sequence of the multispectral wave bands to obtain a multidimensional vector, and taking the multidimensional vector as the current spectrum vector of each pixel point.
3. The method of claim 2, wherein the step of performing flat field correction on the spectral image based on a preset whiteboard image to obtain a corrected spectral image comprises: masking the lens of the imaging component, and collecting an image through the lens to obtain a dark field image; The spectrum image and the dark field image are subjected to difference to obtain a spectrum image with uniform illumination; And carrying out flat field correction on the spectrum image with uniform illumination according to a preset whiteboard image to obtain a corrected spectrum image.
4. The method of claim 1, wherein the step of identifying the current spectral vector by a predetermined contamination detection model to obtain a contamination detection result comprises: Carrying out data construction on the current spectrum vector according to the wave band number of the multispectral lighting component to obtain a data cube containing space and spectrum information; Intercepting the data cube to obtain a spectrum data block with a preset specification; And identifying the spectrum data block through a preset pollution detection model to obtain a pollution detection result of the optical element to be detected.
5. The method of claim 4, wherein the step of identifying the spectrum data block by a predetermined contamination detection model to obtain a contamination detection result of the optical element to be tested comprises: inputting the spectrum data block into a preset pollution detection model, wherein the preset pollution detection model comprises a downsampling layer, a feature extraction layer, an attention layer and an output layer; The downsampling layer is used for downsampling the spectrum data block to obtain a downsampled data block; the feature extraction layer is used for extracting features of the downsampled data block to obtain spatial and spectral joint features; the attention layer is used for carrying out attention weighting on the spatial and spectral combined characteristics to obtain weighted spatial and spectral combined characteristics, and the output layer is used for identifying the weighted spatial and spectral combined characteristics to obtain a pollution detection result.
6. The method of claim 1, wherein the step of acquiring an image of the illuminated optical element under test by the imaging component to obtain a multi-band spectral image further comprises, prior to the step of: acquiring sample spectrum images of various pollutants in an optical element, and extracting features of the sample spectrum images to obtain corresponding sample spectrum vectors; labeling the sample spectrum vector to obtain a pollutant class label; And performing model learning training through the sample spectrum vector and the pollutant class label to obtain a preset pollution detection model.
7. The method according to any one of claims 1 to 6, wherein the step of acquiring an image of the illuminated optical element to be measured by the imaging means to obtain a multi-band spectral image comprises: Calibrating a region to be detected of the optical element to be detected; sequentially polishing the region to be detected according to multispectral wave bands through the multispectral lighting component; And under the illumination of each spectrum band, the imaging component is used for carrying out image acquisition on the illuminated region to be detected, so as to obtain a multi-band spectrum image.
8. A contaminant detection device, the device comprising: The image acquisition module is used for acquiring images of the illuminated optical element to be detected through the imaging component to obtain a multi-band spectrum image, and the multi-spectrum illumination component is used for illuminating the optical element to be detected placed in the detector according to multi-spectrum bands; the feature extraction module is used for extracting features of the spectrum image to obtain a current spectrum vector corresponding to the spectrum image; the model identification module is used for identifying the current spectrum vector through a preset pollution detection model to obtain a pollution detection result, and the preset pollution detection model is a deep learning model obtained based on sample spectrum image training of various pollutants.
9. A detector, characterized in that the detector is provided with a multispectral illumination component and an imaging component; The detector further comprises a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program configured to implement the steps of the contaminant detection method according to any one of claims 1 to 7.
10. A storage medium, characterized in that the storage medium is a computer-readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, carries out the steps of the contaminant detection method according to any one of claims 1 to 7.

Description

Contaminant detection method, device, detector and storage medium Technical Field The present application relates to the field of device detection technologies, and in particular, to a method and apparatus for detecting a contaminant, a detector, and a storage medium. Background Currently, in the precise adjustment process of high-end optical systems (such as assembly and adjustment of an objective lens of a lithography machine, a main lens of a space camera and a resonant cavity lens of a high-energy laser), the surface cleanliness of an optical element is a key factor affecting the final performance. Because micro-and even nano-scale contaminants, such as dust, fibers, skin grease, condensation of silicone oil vapors, saliva splashes, etc., can lead to increased scattering losses of the system, reduced modulation transfer functions, or formation of damage points under intense laser light. Therefore, in order to effectively control the surface cleanliness of the optical element, it is generally necessary to perform rapid contamination detection at the site of adjustment. When the existing pollution screening is carried out in the on-site adjustment link, the manual visual inspection under a high-intensity illuminating lamp or the oblique illumination observation is carried out by assisting a single-wavelength (such as white light or specific color light) light source. For example, the lens surface is illuminated at an angle with an LED flashlight to see if there is a bright spot or hazy area due to contaminant scattering. However, the specific components of the pollutant are difficult to identify by the method, and the situation that the optical film layer is damaged due to incorrect cleaning operation may exist later. Disclosure of Invention The application mainly aims to provide a pollutant detection method, a pollutant detection device, a pollutant detection instrument and a pollutant storage medium, and aims to solve the technical problem that pollutant components are difficult to identify in a traditional single-wavelength oblique-emission inspection optical element mode. In order to achieve the above object, the present application provides a contaminant detection method applied to a detector provided with a multispectral illumination component and an imaging component, the method comprising: the imaging component is used for acquiring images of the illuminated optical element to be detected to obtain a multi-band spectrum image, and the multi-spectrum illumination component is used for illuminating the optical element to be detected placed in the detector according to multi-spectrum bands; extracting features of the spectrum image to obtain a current spectrum vector corresponding to the spectrum image; And identifying the current spectrum vector through a preset pollution detection model to obtain a pollution detection result, wherein the preset pollution detection model is a deep learning model based on sample spectrum image training of various pollutants. In an embodiment, the step of extracting features of the spectral image to obtain a current spectral vector corresponding to the spectral image includes: Performing flat field correction on the spectrum image according to a preset whiteboard image to obtain a corrected spectrum image, wherein the whiteboard image is a reference image obtained by photographing a diffuse reflection standard whiteboard; extracting the relative reflectivity value of each pixel point by taking each pixel point in the corrected spectrum image as a unit; And arranging the relative reflectivity values of each pixel point according to the wavelength sequence of the multispectral wave bands to obtain a multidimensional vector, and taking the multidimensional vector as the current spectrum vector of each pixel point. In an embodiment, the step of performing flat-field correction on the spectrum image according to a preset whiteboard image to obtain a corrected spectrum image includes: masking the lens of the imaging component, and collecting an image through the lens to obtain a dark field image; The spectrum image and the dark field image are subjected to difference to obtain a spectrum image with uniform illumination; And carrying out flat field correction on the spectrum image with uniform illumination according to a preset whiteboard image to obtain a corrected spectrum image. In an embodiment, the step of identifying the current spectrum vector by a preset pollution detection model to obtain a pollution detection result includes: Carrying out data construction on the current spectrum vector according to the wave band number of the multispectral lighting component to obtain a data cube containing space and spectrum information; Intercepting the data cube to obtain a spectrum data block with a preset specification; And identifying the spectrum data block through a preset pollution detection model to obtain a pollution detection result of the optical element to be detecte